"cor" file for problem analysis

[fabricebourdel]

Hello,

I have 2 "sections" in "diagnosis files" that appeared after a database check, one on 29th 01 and the other yesterday.

What can i do with these files ?

****For the "first" core file, i have done 2 checks :

First :

2009-01-29 18:02:18 10874 ERR 51080 SYSERROR -9005 BD Illegal key

2009-01-29 18:03:10 10874 ERR 53000 B*TREE 07010000000000019823000000000000

2009-01-29 18:03:10 10874 ERR 53000 B*TREE Index Root 480166

2009-01-29 18:03:10 10874 ERR 53348 B*TREE bd402SearchIndexForQuali: 481217

2009-01-29 18:03:10 10874 ERR 53250 INDEX Bad Index 480166 (Root)

2009-01-29 18:03:10 10874 ERR 53250 INDEX Reason "System error: BD Invalid invli"

2009-01-29 18:03:11 10874 ERR 51080 SYSERROR -9041 BD Index not accessible

2009-01-29 18:04:28 10876 ERR 53019 CHECK Base error: index_not_accessib

2009-01-29 18:04:28 10876 ERR 53019 CHECK Root pageNo: 480166

2009-01-29 18:04:30 10876 ERR 53000 CHECK Check data finished unsuccessfully

2009-01-29 18:14:58 10873 ERR 53019 CHECK Base error: index_not_accessib

2009-01-29 18:14:58 10873 ERR 53019 CHECK Root pageNo: 480166

2009-01-29 18:14:59 10873 ERR 53000 CHECK Check data finished unsuccessfully

Second :

2009-01-29 18:17:21 9746 ERR 53000 CHECK Check data finished unsuccessfully

2009-01-29 18:29:31 9744 ERR 53000 B*TREE 07010000000000019823000000000000

2009-01-29 18:29:31 9744 ERR 53000 B*TREE Index Root 480166

2009-01-29 18:29:31 9744 ERR 53367 B*TREE bd400_DeleteSubTrees: 481217

• For the "Second" core file, i have done 1 check :

2009-02-05 21:07:10 4672 ERR 53370 B*TREE Illegal record length: 7823

2009-02-05 21:07:10 4672 ERR 53370 B*TREE Corrupted data page: 206842

2009-02-05 21:07:10 4672 ERR 53000 B*TREE 0701000000000001CBDE000000000000

2009-02-05 21:07:10 4672 ERR 53000 B*TREE Index Root 662212

2009-02-05 21:07:10 4672 ERR 53250 INDEX Bad Index 662212 (Root)

2009-02-05 21:07:10 4672 ERR 53250 INDEX Reason "System error: BD Illegal entry"

2009-02-05 21:07:10 4671 ERR 53019 CHECK Base error: index_not_accessib

2009-02-05 21:07:10 4671 ERR 53019 CHECK Root pageNo: 662212

2009-02-05 21:20:02 4671 ERR 53000 CHECK Check data finished unsuccessfully

I guess a disk problem. The disks subsystem is in raid 10 (6 disks, 3 stripped * 2 disks mirrored).

I haven't to now installed specific software to read states about the sas card and the disks, but

i will seach.

I big question : when a "corrupted date page" arrived, how do maxDB handle these problem ?

Does le data in the page be lost after repair ?

[lbreddemann · Accepted solution]

> I have 2 "sections" in "diagnosis files" that appeared after a database check, one on 29th 01 and the other yesterday.

>

> What can i do with these files ?

Since you've to ask the answer is: nothing.

Developers can use the COR files (these are the dumped corrupt pages) and check, why they were found corrupt.

> I big question : when a "corrupted date page" arrived, how do maxDB handle these problem ?

> Does le data in the page be lost after repair ?

MaxDB does not handle corruptions different than any other DBMS for SAP.

It reports that there is something wrong and gives up trying to read the data.

(Unfortunately MaxDB has yet to learn that not all corruptions are a reason to crash - but that's a different topic...).

Corrupted data can never be repaired - by no DBMS available. It may be possible to recreate the data (e.g. Index rebuild, reloading of BW data etc.) but the database software cannot know what was supposed to be in the damaged data page.

That's one of the reasons, why taking and checking database backups is crucial.

Anyhow, if you are a SAP customer, don't miss to open a support call for this.

regards,

Lars

[markus_doehr2]

> 2009-01-29 18:03:10 10874 ERR 53250 INDEX Bad Index 480166 (Root)

> 2009-01-29 18:29:31 9744 ERR 53000 B*TREE Index Root 480166

> 2009-02-05 21:07:10 4672 ERR 53250 INDEX Bad Index 662212 (Root)

> 2009-02-05 21:07:10 4671 ERR 53019 CHECK Root pageNo: 662212

You can check the affected index by doing

select * from roots where root = <pageNo>

If only an Index is affected you can drop and recreate it.

> I big question : when a "corrupted date page" arrived, how do maxDB handle these problem ?

> Does le data in the page be lost after repair ?

If tables are affected you better have a backup, one can't "repair" those problems.

Markus